Paper detail

A structural analysis on Gravity of Trade: on removing distance from the model

The Gravity Model is the workhorse for empirical studies in International Economies for its empirical power and it is commonly used in explaining the trade flow between countries; it relies on a function that relates the trade with the masses of the two countries and the distance (as a proxy of the trasport costs) between them. However, the notion that using of distance functions in conventional interaction models effectively captures spatial dependence in international flows has long been challenged. It has been recently fully recognized that a spatial interaction effect exists essentially due to the spatial spillover and the third country effect. This motivates the introduction of the spatial autoregressive components in the so-called spatial gravity model of trade. A so-called weight matrix is used in order to define the set of the spatial neighbors and it is traditionally based on the inverse of the distance. Two issues follow from this standard procedure: the first regards the biasness of the distance if it is used as a proxy of the transport costs in a panel data, the second is related to the collinearity emerging if we use distance twice. So, several attempt were made in the recent literature having the scope of remove the distance. We propose a theoretically consistent procedure based on Anderson, Van Wincoop derivation model, and some ad-hoc tests, relating to this attempt. The empirical results based on a 22-years panel of OECD countries are conforting, and they allow us to estimate the model without the distance, if properly replaced by a set of fixed effects. This article, in addition, fits in the dispute about how to estimate the multilateral resistance terms.